Split of winners:losers

[BigBaitsim (milo)]

I keep hearing people say that 95% of poker players are long-term losers. My PT data shows a 40/60 split (as does another poster in a different thread). I know that for many there is a very small sample size, but does anyone have any data to back up the 95% number I hear bandied about so often?

[dogmeat]

There is no "data" to back-up most of the claims made on the 2+2 forum, unless they are in regard to specific hands and their chances against a range of other hands. As for the number of losers playing poker every day, I believe the number is 95% when we talk about a lifetime of poker.

As far as Pokertracker goes, you are correct when you mention the sample size. Your hand samples are made up almost entirely of tables that you sat at for an hour or so. When you look at how somebody does over just an hour, many will be winning that have no chance of winnig when you get to 10 hours, or 50 hours. If you sort your hands by only the players that have 100 or more hands against you, you will see some changes. However, this "new" sample will also be flawed, because it will take players that you have played against more than once, and the players that are winning are more likely to be playing more, so your sample is tainted by not being a representative sample of all players.

Dogmeat /images/graemlins/spade.gif

[Rudbaeck]

[ QUOTE ]
As far as Pokertracker goes, you are correct when you mention the sample size. Your hand samples are made up almost entirely of tables that you sat at for an hour or so. When you look at how somebody does over just an hour, many will be winning that have no chance of winnig when you get to 10 hours, or 50 hours.

[/ QUOTE ]

If John has a 95% chance of being a loser over 50 hours, and we make 10,000 clones of John and observe them over an hour each they will be pretty close to 9,500 losers over that hour.

I think the 95% figure is an overstatement. But it's not as low as PokerTracker seems to imply either.

People playing at a limit they are comfortable with both economically and skillwise will most likely not be BIG losers.

And one BIG loser can sustain many, many small winners. (Just think how many high-rolling sharks Bill Gates could create by _only_ steaming off that days profit.)

I think one reason this number is so inflated is that so many poker authorities are in Vegas. And they play high stakes almost exclusively. Here something rather interesting happens, almost everyone they meet will be a loser in Vegas.

But, that everyone could be a life-long small winner. Because a fish in Vegas can still be a shark in Tulsa.

Casino play is off course going to put a serious dent in the number of winning players. If everyone played home games exclusively I'd guess that the ratio of winners to losers would be something like 6:4. But the rake is going to eat away and make a huge group of modest winners into modest losers.

[moondogg]

Does PT include rake in "winners vs. losers"? (I don't think so, but I don't have it in front of me)

If you have 1000 throwing money back and forth on 50/50 shots, on average they will all break even in the end. However, if you start skimming a little here and there, they will all lose.
Theorectically, a game of 10 equally bad fish can play forever with no rake, because none of them have an advantage over each other. With a rake, they would all lose on average, they the expected amount would be exactly rake% * #hands.

I also have a 46:60 ratio in PT, and it has really bothered me for the same reasons. I think the variance for that number would be absolutely huge, just based on the fact that so people in the database have only 30-40 hands. Granted, the ratios should still work out the same, but I would think it would take a hell of a long time to get there.

However, food for though (just thinking out loud): If you sit at a table with 9 fish, you are almost never going to be the only winner. There will always be some at the table that will a little by luck (despite being a long term loser) and others that lose a little (more) by luck. Because you will probably never see any of those people at a table again (if you are playing at Party/Empire), you will be getting a constant feed of 3 or 4 false "winners" every time you sit down at a table with 9 opponents, just because most of your opponents are coin-flipping with each other, and then losing the rake.

Note: this is not an anti-rake post in any way, and I'm not saying rake is inherently bad (or good). It is just statement of the concept that if several people make self-weighting bets on purely random (to them) events, they will both converge to breaking even, unless there is some rake/vig/commission, in which case they convert to breaking even minus the rake.

[Patriarch]

I think that the 95% figure includes the millions of people who might only play poker once or twice in their entire life, and lose.

[TimM]

[ QUOTE ]
If John has a 95% chance of being a loser over 50 hours, and we make 10,000 clones of John and observe them over an hour each they will be pretty close to 9,500 losers over that hour.

[/ QUOTE ]

This is wrong.

First, someone has to be really bad at poker to have a 95% chance of being a loser after 50 hours. But an example of such a person would be someone who loses 35BB/hr on average, with a standard deviation of 15BB. After 50 hours, this person would have about a 95% chance of being in the hole. Guess what his chance of being behind after one hour is. It's a little over 59%, so only 59% of the Johnny clones will be behind after one hour (this is assuming they are not playing each other, but in 10000 separate games).

(I don't know the typical SD of losing players, but note that if you increase their SD, you also have to increase the "loss rate" to give them the same chance of being behind after a certain amount of time. If you give Johnny a -19BB/hr and an 8 SD/hr, you get the same results - ~95% loss after 50 hours, ~59% loss after 1 hour).

If we take something more reasonable, someone with a 95% chance of being behind after 1000 hours, they would still need a win rate of about -8BB/hr, and a SD of 15BB/hr (or -4BB/hr, SD 7.5BB/hr). This player will only have a 52% chance of being behind after 1 hour.

[Nottom]

[ QUOTE ]
If John has a 95% chance of being a loser over 50 hours, and we make 10,000 clones of John and observe them over an hour each they will be pretty close to 9,500 losers over that hour.


[/ QUOTE ]

I realy don't feel like thinking about the math, but my gut tells me this might not be the case.

*after I posted this, I started thinking about how to do the math and was about to comeback and post it and saw that someone already did.

[TimM]

I thought it was an interesting problem, but also didn't feel like doing the math, so I played the game of plug numbers into Excel. I found later that no matter what the win rate and SD, someone who has a 95% chance of losing after 50 hours has a 59.2% chance of losing after one hour, so there must be a way to solve this problem in general, and make the win rate and SD drop out of the equation.

[Moozh]

Imagine this problem.

You take everyone in your PT database and for simplicity you make 2/3 of them losers and 1/3 winners. Say there are 9000 people. That makes 3000 winners and 6000 losers.

Ok, now, repeat the process. Randomly reassign 1/3 winners and 2/3 losers.

What happens?

Of the initial 1/3 winners, 1/3 of them are up double, and 2/3 of them are back to even. And the same for the losers which makes:

1/9 (1000 people) winners for two sessions
2/9 + 2/9 (4000 people) break even
4/9 (4000 people) losers

After two sessions, the split is no longer 1/3 winners, it's much worse. Now do it again.

of the 1/9 winners, 1/3 of them (1/27 total) are 3 time winners, 2/3 of them (2/27) drop to 1 time winners. This works out to:

1/27 (333 people) are up 3
6/27 (2000 people) are up 1
12/27 (4000 people) are down 1
8/27 (2667 people) are down 3

This leaves us with 26% winners and 74% losers even though we started with a 33% 67% split.

Do you see how as this continues, the split becomes greater and greater? The more hands you get on the other people in your database, the more the difference between winners and losers will become apparent.

[Rudbaeck]

Don't tell my old statistics professor.

I'll just go put on the funny hat and stand in the corner for a while.

[alThor]

[ QUOTE ]
I found later that no matter what the win rate and SD, someone who has a 95% chance of losing after 50 hours has a 59.2% chance of losing after one hour, so there must be a way to solve this problem in general, and make the win rate and SD drop out of the equation.

[/ QUOTE ]

Everything you wrote in your two posts is correct. Here is a formula if you want to play with other numbers.

Suppose that after n hours, someone has a p chance of being a loser. Then what is their chance or being a loser after only one hour? You can compute it with this function in Excel:

=NORMSDIST(NORMSINV(p)/SQRT(n))

So typing NORMSDIST(NORMSINV(.95)/SQRT(50)) would yield 0.59197, as you said.

A similar formula would work in the other direction. I leave it as a homework exercise. /images/graemlins/wink.gif

alThor

[mmcd]

I think people aren't accounting for the fact that the "poker economy" is structured like a pyramid. A 3/6 winner will move up to 5/10, 5/10 to 10/20, 10/20 to 20/40, etc. Eventually almost all players get to a game that they just can't beat, and ego will keep them there until they're broke/almost broke. Sure there are a few consistant pros out there who stick within a particular limit and grind out a relatively steady living, but there are also a lot more players who are consistant losers at every level.
Of all the money that is out there "in play", I think the rake takes a decent chunk, and the truly exceptional players (maybe top 1% or less) are going to eventually end up with a large portion of the rest.

What keeps it going is that the bad players, and theres a LOT of them, are willing to lose a LOT of money on a consistant basis.

[jdl22]

Mine too is roughly 40/60.

[Blarg]

Online poker playing is very unrepresentative of what poker traditionally has been and still often is, too.

Poker in casino now and in general historically has been something that is very social, and that naturally goes along, for most people by far, with having a good time, letting it all hang out - smoking, drinking, whoring, even taking drugs for some.

The Friday night payday poker player and the weekend partying poker goofball aren't the type you see online. People online may be unskilled, some may even be stupid, but they're not at all in the social situations which lead to so much show-off behavior at live tables, and booze-impaired judgment.

The 5% winners numbers seems perhaps even generous for live play. For online play, I'm not sure what the number is, but think it's notably better than it is in casinos. So Pokertracker will also at best show only part of the grand loserdom that is most of poker players.

[ctv1116]

Consider the following situation: we play in a game where every player loses at -1BB/100 with a standard deviation of 15BB/100. Now let all the players play 100 hands, which probably is the number hands in the average PT database. What will be the win/loss ratio for this population of players, given the normal distribution? It will be about 45/55. In reality, by definition, ALL of the players are long-term losers, yet in a theoretical PT database, the winners:losers ratio is NOT 0:100